Repository files navigation

Predictive Horizontal Pod Autoscalers (PHPAs) are Horizontal Pod Autoscalers (HPAs) with extra predictive capabilities,allowing you to autoscale using statistical models for ahead of time predictions.

PHPAs can better scaling results by making proactive decisions to scale up ahead of demand, meaning that aresource does not have to wait for performance to degrade before autoscaling kicks in.

Any systems that have regular/predictable demand peaks/troughs.

Some use cases:

• A service that sees demand peak between 3pm and 5pm every week day, this is a regular and predictable load whichcould be pre-empted.
• A service which sees a surge in demand at 12pm every day for 10 minutes, this is such a short time interval thatby the time a regular HPA made the decision to scale up there could already be major performance/availablity issues.

PHPAs are not a silver bullet, and require tuning using real data for there to be any benefits of using it. A poorlytuned PHPA could easily end up being worse than a normal HPA.

This project works by doing the same calculations as the Horizontal Pod Autoscaler does to determine how many replicasa resource should have, then applies statistical models against the calculated replica count and the replica history.

The minimum Kubernetes version the autoscaler can run on isv1.23 because it relies on theautoscaling/v2 API whichwas only available inv1.23 and above.

The autoscaler is only tested against the latest Kubernetes version - if there are bugs that affect older Kubernetesversions we will try to fix them, but there is no guarantee of support.

• Functionally identical to Horizontal Pod Autoscaler for calculating replica counts without prediction.
• Choice of statistical models to apply over Horizontal Pod Autoscaler replica counting logic.
  • Holt-Winters Smoothing
  • Linear Regression
• Allows customisation of Kubernetes autoscaling options without master node access. Can therefore work on managedsolutions such as EKS or GCP.
  • CPU Initialization Period.
  • Downscale Stabilization.
  • Sync Period.

PHPAs are designed to be as similar in configuration to Horizontal Pod Autoscalers as possible, with extraconfiguration options.

PHPAs have their own custom resource:

apiVersion:jamiethompson.me/v1alpha1kind:PredictiveHorizontalPodAutoscalermetadata:name:simple-linearspec:scaleTargetRef:apiVersion:apps/v1kind:Deploymentname:php-apacheminReplicas:1maxReplicas:10behavior:scaleDown:stabilizationWindowSeconds:0metrics:    -type:Resourceresource:name:cputarget:averageUtilization:50type:Utilizationmodels:    -type:Linearname:simple-linearlinear:lookAhead:10000historySize:6

This PHPA acts like a Horizontal Pod Autoscaler and autoscales to try and keep the target resource's CPU utilization at50%, but with the extra predictive layer of a linear regression model applied to the results.

The operator for managing Predictive Horizontal Pod Autoscalers can be installed using Helm:

VERSION=v0.13.2HELM_CHART=predictive-horizontal-pod-autoscaler-operatorhelm install${HELM_CHART} https://github.com/jthomperoo/predictive-horizontal-pod-autoscaler/releases/download/${VERSION}/predictive-horizontal-pod-autoscaler-${VERSION}.tgz

Check out thegetting startedguide and theexamples for ways to use Predictive Horizontal Pod Autoscalers.

See thewiki for more information, such as guides andreferences.

See theexamples/ directory for working code samples.

Developing this project requires these dependencies:

• Go >=1.20
• Python ==3.8.x
• Helm ==3.9.x

Any Python dependencies must be installed by running:

pip install -r requirements-dev.txt

This extensively uses the thejthomperoo/k8shorizmetrics libraryto gather metrics and to evaluate them as the Kubernetes Horizontal Pod Autoscaler does.

It is recommended to test locally using a local Kubernetes managment system, such ask3d (allows running a small Kubernetes cluster locally using Docker).

You can deploy a PHPA example (see theexamples/ directory for choices) to test your changes.

• make run - runs the PHPA locally against the cluster configured in your kubeconfig file.
• make docker - builds the PHPA image.
• make lint - lints the code.
• make format - beautifies the code, must be run to pass the CI.
• make test - runs the unit tests.
• make doc - hosts the documentation locally athttps://localhost:8000.
• make coverage - opens up any generated coverage reports in the browser.